From ea5d519358d12cb6f9bb9bcdc0b45b5c091937d8 Mon Sep 17 00:00:00 2001
From: crusaderky <crusaderky@gmail.com>
Date: Thu, 21 Sep 2023 13:52:13 +0100
Subject: [PATCH] XFAIL test_dataframe_cow_chain

---
 tests/benchmarks/test_dataframe.py |  6 +++++-
 tests/benchmarks/test_parquet.py   | 13 ++-----------
 tests/utils_test.py                |  8 ++++++++
 3 files changed, 15 insertions(+), 12 deletions(-)

diff --git a/tests/benchmarks/test_dataframe.py b/tests/benchmarks/test_dataframe.py
index 6c3d032708..bdc1cc9409 100644
--- a/tests/benchmarks/test_dataframe.py
+++ b/tests/benchmarks/test_dataframe.py
@@ -1,7 +1,8 @@
+import pytest
 from dask.sizeof import sizeof
 from dask.utils import format_bytes
 
-from ..utils_test import cluster_memory, timeseries_of_size, wait
+from ..utils_test import HAS_PYARROW12, cluster_memory, timeseries_of_size, wait
 
 
 def print_dataframe_info(df):
@@ -69,6 +70,9 @@ def test_filter(small_client):
     wait(result, small_client, 10 * 60)
 
 
+@pytest.mark.xfail(
+    HAS_PYARROW12, reason="https://github.com/coiled/benchmarks/issues/1002"
+)
 def test_dataframe_cow_chain(small_client):
     memory = cluster_memory(small_client)  # 76.66 GiB
 
diff --git a/tests/benchmarks/test_parquet.py b/tests/benchmarks/test_parquet.py
index 383078caea..cbf57f040a 100644
--- a/tests/benchmarks/test_parquet.py
+++ b/tests/benchmarks/test_parquet.py
@@ -12,17 +12,9 @@
 import pandas
 import pytest
 from coiled import Cluster
-from packaging.version import Version
 
 from ..conftest import dump_cluster_kwargs
-from ..utils_test import run_up_to_nthreads, wait
-
-try:
-    import pyarrow
-
-    HAS_PYARROW12 = Version(pyarrow.__version__) >= Version("12.0.0")
-except ImportError:
-    HAS_PYARROW12 = False
+from ..utils_test import HAS_PYARROW12, run_up_to_nthreads, wait
 
 
 @pytest.fixture(scope="module")
@@ -51,8 +43,7 @@ def parquet_client(parquet_cluster, cluster_kwargs, upload_cluster_dump, benchma
 
 
 @pytest.mark.xfail(
-    HAS_PYARROW12,
-    reason="50x slower than PyArrow 11; https://github.com/coiled/benchmarks/issues/998",
+    HAS_PYARROW12, reason="https://github.com/coiled/benchmarks/issues/998"
 )
 @run_up_to_nthreads("parquet_cluster", 100, reason="fixed dataset")
 def test_read_spark_generated_data(parquet_client):
diff --git a/tests/utils_test.py b/tests/utils_test.py
index 8dcd94a952..2397d8b2c8 100644
--- a/tests/utils_test.py
+++ b/tests/utils_test.py
@@ -12,6 +12,14 @@
 from dask.datasets import timeseries
 from dask.sizeof import sizeof
 from dask.utils import format_bytes, parse_bytes
+from packaging.version import Version
+
+try:
+    import pyarrow
+
+    HAS_PYARROW12 = Version(pyarrow.__version__) >= Version("12.0.0")
+except ImportError:
+    HAS_PYARROW12 = False
 
 
 def scaled_array_shape(